Physics-based data-driven simulation of debinding failure and sintering deformation in metal binder jetting

Abstract

Debinding failure, sintering distortions are two of the main challenges in manufacturing of intact and dimensionally precise components with metal binder jetting technology. In the current research, the challenges are addressed by developing physics-based data-driven numerical simulations. Using thermos-optical observations it is demonstrated that debinding failure occurs when the binder is burnt-off from the green parts which lead to the strength reduction of the components. Through a design of experiment on self-stressing samples, it is resulted that stress on the specimens and the print direction are important factors which may affect debinding failure. A simulation-based Weibull model is developed to consider stress and print direction of the samples to predict he failure probability of green parts during debinding. The collected data from the design of experiment are used to fit the parameters of the simulation-based Weibull model. Validation evidence shows that the developed model is able to predict the failure probability of green parts. Furthermore, the model is detect the failure-prone zones with an acceptable accuracy. The creep-like behavior of sintering deformations can be captured by a rate-dependent material model. Therefore, a phenomenological thermo-elasto-viscoplastic material model in the scheme of finite element method can simulate the thermal, elastic and viscoelastic deformations of sintering bodies in a component scale. The aforementioned material model is calibrated for metal binder jetting technology with dilatometry, Dynamic Differential Calorimetry, Laser Flash Analysis, metallographic analysis, and thermos-optical measurements. The validation studies reveals that without considering the characteristics of MBJ green parts such as density variations and anisotropy, the required industrial-acceptable tolerances cannot be achieved by simulations. Using a data-driven approach and with a wider range of experimental data, the material model is developed in a generic format with the ability to adopt for new manufacturing parameters. The experimental study enabled the estimation of porosity-dependent anisotropy, thermal expansion coefficient, and viscosity. Moreover, with the data-driven approach, the green part properties such as density variations and dimensions can be predicted and fed into the material model. The predicted deformations from the data-driven material model have a good agreement with the validation data. Taking advantage of the data-driven numerical simulation and by introducing a compensation algorithm, sintering deformations of a given geometry are iteratively calculated and reversed to prepare a compensated configuration of the geometry. The presented approach results in manufacturing of MBJ parts with international tolerance grade of 14. Moreover, an experimental-based compensation technique is developed which takes use of previously calculated deformation vectors from numerical simulations and scan data of as-sinter configuration. Measuring the deviation between the as-sinter and the target geometry with the help of the normalized deformation vectors from simulations considers the physics of deformation and enables manufacturing of MBJ parts with a higher accuracy, achieving an international tolerance grade of 13.